The Pattern Nobody Wanted to See
I spent most of 2024 integrating Claude's API into a document processing tool for a Bogotá-based startup. Mid-integration, I remember thinking: this is almost boring now. The technology works. It's stable. We're past the hype phase. Then I read about someone allegedly throwing a Molotov cocktail at Sam Altman's house because they feared AI would cause human extinction.
The fear isn't new. But the violence is different.
What struck me wasn't the attack itself—it was the pattern. A 20-year-old with extinction anxiety. An Indianapolis councilman receiving 13 bullets for supporting a data center. Another alleged attack days later. These aren't random acts. They're symptoms of something festering beneath the surface of the AI industry.
Building Products While the World Worries
Here's the uncomfortable part: I build AI tools. I think they're useful. I also understand—genuinely understand—why some people look at what we're building and see an existential threat. Not because I agree entirely, but because I'm not sure I fully disagree either.
- When you're actually integrating language models into production systems, you see the capabilities. They're real.
- You also see the limitations, the hallucinations, the weird failure modes that shouldn't exist but do—this part gets less media coverage
- But you can't unsee what the technology might become
I'm not sure this is the right way to frame it, but here's what I think: the people committing violence against AI leaders aren't primarily motivated by having read research papers. They're motivated by a deeper fear that nobody in the industry has adequately addressed. We've been too busy shipping features to acknowledge the legitimate questions.
The Industry's Silence Problem
OpenAI, Anthropic, Google's DeepMind division—they publish safety research. They genuinely do. But there's a gap between publishing a technical whitepaper and actually communicating with people who are terrified. A massive gap.
When I talk to developers in Bogotá who are just starting with AI, they ask different questions than San Francisco engineers. They ask: who owns this? Who decides what it does? What happens when someone else's government weaponizes it? These aren't paranoid questions. They're practical ones.
The attacks on Altman exist in a world where transparency has been... selective. OpenAI's governance structure has been questioned. The compute requirements for training modern models raise genuine environmental and resource questions. Whether you think these concerns are overblown or not, they're real to the people asking them.
What This Means for People Actually Building
I'm not going to pretend I have answers here. I don't. But I know what I'm doing differently:
- Being honest about what the tools actually do versus the marketing
- Thinking about downstream effects when designing features. Not as a checkbox—actually thinking. This is harder than it sounds, much harder, and I fail at it regularly
- Talking to people who are skeptical instead of just talking to other builders
The violence is indefensible. Full stop. But treating these attacks as aberrations rather than signals would be a mistake. They're not coming from nowhere.
The Unresolved Part
Here's what I genuinely don't know: how do we move forward? Do we slow down? Accelerate? Regulate harder? I've heard passionate arguments for all three, and they all contain something true and something dangerous. The person who threw a Molotov cocktail didn't do it because they read a balanced analysis of AI governance models.
They did it because they felt unheard. That feeling—whether justified or not in its extremism—is worth understanding instead of dismissing.
The AI industry needs to stop acting like skepticism is something to overcome and start treating it as something to engage with. Not to shut down, but to understand. Sam Altman didn't deserve to be attacked. And the fears driving someone to commit violence deserve to be taken seriously, even if the response to those fears absolutely does not.
That's the warning. Not that AI is dangerous in the way the attacker feared. But that ignoring what people fear tends to produce exactly these kinds of outcomes.
